From Tunisian Spices to Icelandic AI: Safa Jemay on Bridging Worlds

From Tunisia to Iceland: A Story That Begins with a Summer Internship

Some of the best conversations I have had on Temjum Taeknina begin with a story I did not expect. When Safa Jemay sat down with me, I thought we would talk mostly about AI consulting. Instead, the conversation opened with the scent of handmade Tunisian spices and the sound of someone learning to roll Icelandic Rs in record time.

Safa arrived in Iceland in 2017, coming from Tunisia for a summer internship in Hveragerdi. She was 23 years old, had never been to the Nordics, and had no particular plan to stay. But something about this small island at the edge of the Arctic caught hold of her. She fell in love with the country and made the decision that would define the next chapter of her life: she would stay, and she would do it properly.

Doing it properly, in Safa's case, meant learning Icelandic in five months, well enough to enroll in a computer science degree at the University of Iceland, taught entirely in Icelandic. If you have ever tried to learn Icelandic, you know this is not a casual achievement. It is a language that has barely changed since the medieval sagas, with grammatical cases and conjugations that trip up even dedicated students for years. Safa did it in under half a year, driven by what she describes simply as determination. Listening to her tell the story, I believed every word.

Bridging Cultures Through Food and Technology

One of the things that struck me most about Safa is how she builds bridges between worlds that seem far apart. Her first business, Mabruka, is a direct expression of this. Her mother, back in Tunisia, makes spice blends by hand from scratch, traditional recipes passed through the family. Safa imports them to Iceland and sells them in major grocery chains like Hagkaup and Kronan. It is a family enterprise in the truest sense, connecting her Tunisian roots to her Icelandic life through something as universal and intimate as food.

There is something deeply human about the idea that a jar of handmade harissa or cumin blend can be a cultural ambassador. When Icelanders reach for one of Mabruka's products in the grocery aisle, they are participating in a tiny act of cultural exchange. Safa does not frame it this grandly herself; she is too practical for that. But the underlying thread is unmistakable. She sees connections where others see distance.

That same instinct for connection led to her second business. After trips to tech conferences in Denmark and Helsinki, colleagues and contacts kept pointing out the obvious: you are already working with AI, why not offer it as a service? And so Vikonnekt was born, an AI and technology consulting firm focused on helping Icelandic businesses and institutions understand and adopt artificial intelligence in practical, grounded ways.

The AI Consulting Journey: Training Before Tools

What I appreciate about Safa's approach to AI consulting is that it starts with people, not software. She has developed what she calls the Cycle of Learning methodology: begin with training and education, move into hands-on workshops, then map the specific challenges an organization actually faces, build a proof of concept, and only then think about scaling. It is a patient, structured approach in a field where many vendors want to skip straight to selling expensive enterprise solutions.

Her ratio is telling: 20 percent education, 80 percent training. The distinction matters. Education is knowing what AI is. Training is knowing how to use it in your specific context, with your specific data, for your specific problems. Too many organizations invest heavily in awareness without ever bridging the gap to applied skill. Safa is focused squarely on that bridge.

"People need to understand the tool before they can trust it. And they need to trust it before they will use it well."

We also touched on the rapidly shifting economics of AI. The arrival of DeepSeek has changed the conversation about what large language models cost to run. Safa mentioned an anecdote about Google sending a user a 37-dollar invoice for heavy Gemini usage, a sign that the era of unlimited free access to powerful AI is evolving. For small economies like Iceland, this cost question is not academic. It shapes what tools are accessible and what kind of local AI infrastructure might be worth investing in. She pointed to Poland as an example: a country investing in building local language models rather than relying entirely on American platforms.

The Bias Problem: When AI Reflects Our Worst Patterns

One of the most important parts of our conversation centered on AI bias. Safa referenced a Bloomberg study that examined how AI systems can reproduce and even amplify racial biases present in their training data. This is not a new finding, but it remains one that too many organizations brush past in their eagerness to adopt AI tools.

Safa brings an important perspective to this discussion. As a Tunisian woman working in Icelandic tech, she has personal experience of what it means to exist at the intersection of multiple identities in a predominantly homogeneous society. When she talks about AI bias, it is not abstract theory. It is something she understands from the inside out. That combination of technical expertise and lived experience makes her voice particularly valuable in a conversation that too often stays within the walls of research labs.

The point she makes is straightforward but essential: if the data going into an AI system carries historical inequities, the outputs will carry them too, often invisibly. Organizations adopting AI need to understand this not as a technical footnote but as a foundational design consideration.

Shadow AI and Organizational Readiness

Perhaps the most practically urgent topic we discussed was Shadow AI, the phenomenon of employees using AI tools that their organizations do not officially know about, have not vetted, and have no policies for. It is happening everywhere. People are using ChatGPT to draft emails, summarize meetings, write reports, even process sensitive data, all without their IT departments having any visibility.

Safa shared a striking statistic: one of the top use cases for ChatGPT globally is people using it as a therapist or psychologist. Sit with that for a moment. Millions of people are sharing their most intimate thoughts and struggles with a language model, often without any understanding of where that data goes or how it might be used. The implications for organizational data are similarly sobering. If employees are pasting internal documents into consumer AI tools, the security and privacy questions are enormous.

The solution, as Safa sees it, is not to ban AI use. That ship has sailed. The solution is to get ahead of it with clear policies, proper training, and sanctioned tools that meet security and ethical standards. This is where her Cycle of Learning methodology becomes not just nice to have but genuinely necessary. Organizations that ignore Shadow AI are not avoiding risk; they are simply choosing not to see it.

Passion as the Engine of Innovation

Throughout our conversation, one theme kept surfacing in different forms: the relationship between passion, freedom, and meaningful work. Safa's daily routine starts with Pilates at 6 AM before heading straight into her working day, a rhythm that speaks to someone who has integrated discipline and purpose into her life in a way that feels natural rather than forced.

"This is not work, this is passion."

That line stayed with me. It connects to something I explored in an earlier episode with Jonatan Solon, the link between entrepreneurship, creative freedom, and the conditions that allow people to do their best thinking. Safa embodies this connection. Running two businesses simultaneously, one rooted in family tradition and the other in cutting-edge technology, requires a kind of energy that cannot be manufactured by salary alone. It comes from genuine engagement with the work itself.

We also drifted into a conversation about universal basic income, viewed through this lens of creative freedom. If people had a baseline of economic security, would more of them build things like Mabruka or Vikonnekt? Would more people take the risk of starting something new? It is a question without a simple answer, but Safa's own story is a compelling data point. When someone has the determination and the breathing room to pursue what matters to them, remarkable things can happen.

Closing Reflection: The View from Two Worlds

Safa Jemay is one of those people who makes you reconsider the boundaries between categories. She is an immigrant and a local. She is a food entrepreneur and a tech consultant. She is academically trained and street-smart. She brings Tunisia to Iceland and Iceland back to Tunisia, through spice jars and slide decks alike.

What I take away from this episode is a reminder that the most interesting people in the AI space are often not the ones who have spent their entire careers in technology. They are the ones who come to it from unexpected angles, carrying experiences and perspectives that the field desperately needs. When Safa talks about AI bias, she is not citing papers she has read. She is describing a world she navigates. When she builds training programs for Icelandic organizations, she draws on the same grit that got her through Icelandic grammar in five months.

In a field that often rewards narrow specialization, Safa is a compelling argument for breadth. For bringing your whole self, all your worlds, all your languages, all your spice blends, to the table. The technology is only as good as the people shaping it, and the people shaping it are only as good as the range of experience they bring.

If you want to hear the full conversation, including Safa's thoughts on local AI infrastructure, the Polish model for language technology, and why she believes the future of AI adoption is about people first and tools second, listen to Season 1, Episode 6 of Temjum Taeknina wherever you get your podcasts.